Implicit conversion from float -0.4017857142857082 to int loses precision